Google’s Gemini 3 Deep Think debuts for Ultra users

Deep Think incorporates the code that won the math olympiad, and is reserved for heavy lifting.
Gemini 3 Deep Think brings some serious AI muscles to tackle the toughest problems. (Picture: Google)
Google just launched it’s most capable model to it highest paid tier, ready to take on the most confounding probblems.

It incorporates the solutions that won gold at the International Mathematical Olympiad and beat the ICPC coding contest, and carries on its duties doing parallel reasoning, letting it try several approaches to a problem at once.

Not for everyone
It is «designed to tackle complex math, science and logic problems that challenge even the most advanced state-of-the-art models,» Google says on its launch page.

Scoring some 41% on Humanities Last Exam and 45.1% on ARC-AGI-2, more than doubling GPT-5.1s score, it also has world-beating performance in some of the most popular benchmarks, and will be hard to beat for competitors.

Read more: Google’s launch page, writeup on 9to5Google. Discussion on r/Singularity.